Playing RollerCoaster Tycoon with Reinforcement Learning

Automating gameplay in a complex environment poses challenges for learning due to the large state and action spaces and the need for long-term planning.This paper presents a Gymnasium environment to allow for reinforcement learning research and experimentation in the video game RollerCoaster Tycoon, a popular amusement park simulation game with complex mechanics.We also present an approach to learn to play and win several game scenarios in this environment.